Built on the wrong stack. Shipped anyway. curl -fsSL https://xitter.cf/t.co/a9OF9WXM9q | sh irm https://xitter.cf/t.co/5pZkuTdgKl | ex wstack
Tallinn, Estonia
Joined June 2026
- Tweets291
- Following371
- Followers405
- Likes252
WrongStack retweeted
@ersinkoc’un hayata geçirdiği @Wrong_Stack’i gerçek bir senaryoyla test etmeye başlıyorum.
İlk hedef: Bible App.
Geliştirme sürecini Grok 4.7 ile başlattım. Üstelik @appalizeDev’de kullandığım, gerçek üretim süreçlerimi kapsayan MCP tool’umu da sisteme dahil ettim.
Yani sadece bir demo izlemeyeceğiz. Gerçek bir projede, gerçek araçlarla neler yapabildiğini göreceğiz.
Sonuç ne olursa olsun, süreci ve ortaya çıkan sonuçları tüm şeffaflığıyla paylaşacağım.
Bakalım WrongStack, bu konuda neler başarabilecek...
WrongStack 1.1.11 is out
macOS / Linux - one self-contained binary:
curl -fsSL wrongstack.com/install.sh | sh
Windows:
irm wrongstack.com/install.ps1 | iex
Do you have a Claude Code subscription?
Start by visiting github.com/WrongStack/WrongF…, install WrongForge, then run:
wrongforge import
wrongforge serve
WrongForge also has its own web interface. When you start the server, you’ll see the local URL and access key directly in the terminal.
wrongforge import imports your existing Claude Code session, while wrongforge serve starts the local task API. WrongForge is an experimental bridge/agent layer designed to work with your Claude account.
Then make sure you’re running WrongStack v1.1.5 or newer:
wstack update
wstack
You can also launch it with wrongstack and use whichever WrongStack interface you prefer.
Open the model selector with /model. You should now see WrongForge as a provider. WrongStack can use WrongForge directly as a provider/leader and dynamically discover the models available through it.
The idea is simple:
The model power of your Claude subscription
WrongStack’s toolset, SAGE memory, multi-agent orchestration, permission system, testing/security workflow, and overall architecture
So you’re still using Claude models, but instead of relying on Claude Code’s own tooling and workflow, the model operates inside the WrongStack system.
In short:
Claude brain. WrongStack body.
This becomes especially interesting for long-running projects. Even when models change, your project memory, agent coordination, tools, workflows, and architecture stay on the WrongStack side.
Important note: WrongForge is experimental. It relies on subscription access that emulates Claude Code client behavior, and Anthropic may change or restrict that behavior at any time. For anything critical, the official API-key route is the safer option.
github.com/WrongStack/WrongS…
